Currently we calculate performance profile based solely on number of cpus used by compute. This results in lower limits that we could potentially give, when nbs is running with extended cpu set (e.g. on amd platforms etc)
Current limits were calculated with assumption nbs is running on 16 cores. We need to get current cpu set from procfs and adjust performance profile accordingly
Тут еще нужно смотреть на пропускную способность сети т.к. нельзя выдать квоты на VM больше чем вообще сеть тянет.
А на хосте еще могут быть и другие пользователи.
https://github.com/ydb-platform/nbs/blob/main/cloud/blockstore/libs/daemon/common/config_initializer.cpp#L168
Currently we calculate performance profile based solely on number of cpus used by compute. This results in lower limits that we could potentially give, when nbs is running with extended cpu set (e.g. on amd platforms etc)
Current limits were calculated with assumption nbs is running on 16 cores. We need to get current cpu set from procfs and adjust performance profile accordingly